as it happens the code shown was pretty transient anyway. For the module test suite that I wrote the code for, I replaced it with:

    my $root1 = HTML::TreeBuilder->new ();
    my $root2 = HTML::TreeBuilder->new ();

    $root1->parse_content ($rendered)->elementify ()
      ->delete_ignorable_whitespace ();
    $root2->parse_content ($expected)->elementify ()
      ->delete_ignorable_whitespace ();

    is ($root1->as_HTML (undef, '   ', {}),
        $root2->as_HTML (undef, '   ', {}), $testName);
[download]

in any case so that I'd get better diagnostics (I see the two HTML fragments when the test fails). However, with a little tweaking to give a traceback the original code would be even better in the test context because it would highlight the difference by reducing the clutter. That version might almost be worth generating a module for.